RDF predicates

A first look at the data and some basic aggregate count later, we see what are the most popular predicate on webages.

• rdf:type is the most common and that make sense! 15-25% estimate
• vcard addresses seems more comon than schema.org addresses
• Schema.org has many Address predicates
• … Take a look

rdf:types Objects

Similarly, we can observe the rdf:type predicate and see what types of objects it describes the most. using filtering we can refine our search

• The fields let you parameterize the query.
• Fields are using SQL 'LIKE' syntax, please also keep the quotes
• …

We look for predicate that describes Nantes with as much uniqness as possible

Basic filtering

Applied to the whole data in a single read.
We search for a few post code matching the city and the city name inside the NQuads.
We filter the search on some very specific predicate.

Sibligns

Here comes the issue, to find any form of neighbourhood we need to join the data to the original dataset. if its a small set then we can broadcast but here is the general method.

Parent / children

Capturing the parent node enable a second row much faster. Same for eventual children

Graph and Subject

A problem of the raw input is that it does not distinguish between subjet String, IRI, or blank node.
Therefore, the join need to be done using both columns. We can here benefit from having sorting it by graph

After we've grouped our data together, we begin to see patterns but also noise. more importantly, Entities remain as different rows and we would like to assemble them together.

In order to find the address card we need to associate the predicates representing an address together. we use pivoting to do so since we already have already extracted the data